JP6831420B2 - A method for evaluating self-driving car trajectory candidates - Google Patents

Description

The present application mainly relates to the operation of an autonomous vehicle. Specifically, the present application relates to a method for evaluating a trajectory candidate of an autonomous vehicle (ADV).

A vehicle traveling in autonomous driving mode (eg, driverless) can free the occupants, especially the driver, from some driving roles. When the vehicle is driven in the autonomous driving mode, it is possible to drive the vehicle when there is the least human-machine interaction or when there are no passengers by navigating to each position using the in-vehicle sensor. Can be allowed.

The ADV can perform self-guidance based on the driving locus. The driving locus can be divided into a vertical component and a horizontal component. The longitudinal component or longitudinal locus is a vehicle behavior in which the vehicle travels in the longitudinal direction along a predetermined route (for example, a station path). The vertical locus can determine the position, velocity and acceleration of the ADV at a predetermined timing. Therefore, longitudinal trajectory generation is an essential component of a semi-autonomous or fully autonomous vehicle.

Several factors need to be considered in the process of generating the longitudinal locus. These factors may be factors for the safety, comfort and traffic rule compliance of passengers and / or nearby pedestrians in ADV vehicles. In order to realize safe operation of ADV, it is necessary to consider obstacles in the surrounding environment in the trajectory generation process. For comfortable operation, the trajectory needs to be smooth and effective. That is, it may be a locus having a stable acceleration capable of driving the ADV from the current position to the destination within a reasonable time. Finally, the trajectory must follow local traffic rules, etc., that is, stop at a red traffic light and a stop sign. With all these factors in mind, there is a lack of an effective way to generate trajectories.

One aspect of the present application relates to a computer implementation method for determining a trajectory for operating an autonomous vehicle (ADV). The method is used in an autonomous vehicle (ADV) to generate a plurality of trajectory candidates from a start point to an end point of a specific driving scene, and to the current state of the autonomous vehicle and the end point related to the start point. Based on the end state of the related autonomous vehicle, the step of generating the reference locus related to the target corresponding to the driving scene and the locus candidate are compared with each other. , A step of generating a target cost showing similarity between the locus candidate and the reference locus, and a locus candidate as a target locus for driving the autonomous vehicle based on the target cost of the locus candidate. Includes a step of selecting any one of them.

Another aspect of the present application relates to a non-transitory machine-readable medium in which the directive is stored. When the command is executed by one or more processors, the one or more processors are made to execute the process. The process generates a plurality of trajectory candidates from the start point to the end point of a specific driving scene used in the autonomous driving vehicle (ADV), and is related to the current state of the autonomous driving vehicle related to the start point and the end point. Based on the end state of the autonomous driving vehicle, a reference locus related to the target corresponding to the driving scene is generated, and for each of the locus candidates, the locus candidate and the reference locus are compared with the locus candidate. A target cost showing similarity to the reference locus is generated, and one of the locus candidates is selected as the target locus for driving the autonomous driving vehicle based on the target cost of the locus candidate. Including that.

Another aspect of the present application relates to a data processing system comprising one or more processors and a memory connected to the one or more processors to store instructions. When the command is executed by the one or more processors, the one or more processors are made to execute the process. The process includes a step of generating a plurality of trajectory candidates from the start point to the end point of a specific driving scene used in the autonomous driving vehicle (ADV), and the current state of the autonomous driving vehicle and the end point related to the start point. Based on the end state of the related autonomous vehicle, the step of generating the reference locus related to the target corresponding to the driving scene and the locus candidate are compared with each other. , A step of generating a target cost showing similarity between the locus candidate and the reference locus, and a locus candidate as a target locus for driving the autonomous vehicle based on the target cost of the locus candidate. Includes a step of selecting any one of them.

Embodiments of the present application are shown in non-limiting, but exemplary, embodiments in each of the drawings, the same drawing reference numerals in the drawings indicating similar elements.

It is a block diagram which shows the network system by one Embodiment.

It is a block diagram which shows the example of the autonomous driving vehicle by one Embodiment.

FIG. 6 is a block diagram illustrating an example of a sensing / planning system used with an autonomous vehicle according to an embodiment. FIG. 6 is a block diagram illustrating an example of a sensing / planning system used with an autonomous vehicle according to an embodiment.

It is a block diagram which shows the example of the locus generation module by one Embodiment.

It is a block diagram which shows the exemplary scene of ADV by one Embodiment.

It is an example which shows the speed-time graph of a plurality of trajectories by one Embodiment.

It is an example which shows the station-time graph of a plurality of trajectories according to one embodiment.

It is a block diagram which shows the example of the locus evaluation module by one Embodiment.

It is an example which shows the station-time graph of the locus candidate by one Embodiment.

It is an example which shows the reference velocity curve by some embodiments. It is an example which shows the reference velocity curve by some embodiments. It is an example which shows the reference velocity curve by some embodiments.

It is an example which shows the speed-time graph of the target evaluation by one Embodiment.

It is a flowchart which shows the method executed by ADV by one Embodiment.

It is a block diagram which shows the data processing system by one Embodiment.

Hereinafter, various embodiments and embodiments of the present application will be described with reference to the details of the description, and the drawings show the various embodiments. The following description and drawings are intended to illustrate the present application and should not be construed as limiting. Many specific details are described in order to fully understand the various embodiments of the present application. However, in some cases, the details of well-known or prior art are not described in order to provide a brief description of the embodiments of the present application.

As used herein, the term "one embodiment" or "embodiment" means that a particular feature, structure or property described in combination with that embodiment may be included in at least one embodiment of the present application. .. The phrase "in one embodiment" does not necessarily refer to the same embodiment throughout this specification.

According to one aspect, the system generates a plurality of trajectory candidates from the start point to the end point of a specific driving scene used in an autonomous vehicle (ADV). The system generates a reference locus corresponding to the driving scene based on the current status of the ADV related to the starting point and the ending status of the ADV related to the ending point. Here, the reference trajectory is related to the target. For each of the locus candidates, the system compares the locus candidate with the reference locus and generates a target cost that indicates the similarity between the locus candidate and the reference locus. The system selects any one of the locus candidates as the target locus for driving the ADV based on the target cost of the locus candidate.

According to one embodiment, when comparing each of the locus candidates with the reference locus, the locus candidates are divided into a plurality of candidate sections. The reference locus is divided into a plurality of reference sections. For each of the locus sections, determine the velocity difference between the speed of the candidate section and the corresponding reference section. That is, the speed of the candidate section is compared with the speed of the reference section corresponding to the same timing. Based on the speed difference between the candidate interval and the associated reference interval, the target cost is calculated, for example, by a predetermined algorithm. In a specific embodiment, the target cost is calculated by adding the square of the speed difference between the candidate section and the reference section.

According to other aspects, for each of the locus candidates, the most different obstacles are recognized in the locus candidates. Measure the distance between the trajectory candidate and the obstacle. Based on the distance between the trajectory candidate and the obstacle, the safety cost associated with the trajectory candidate is calculated, for example, by a safety cost algorithm or function. Then, the total trajectory cost of the trajectory candidate is calculated based on the target cost and the safety cost. A target locus is selected from the locus candidates based on the total locus cost (for example, the minimum total locus cost).

According to the other aspect, for each of the locus candidates, the rate of change of the acceleration along the locus candidate is determined. The rate of change of acceleration can be determined by the derivative of the velocity along the trajectory candidate. Then, based on the rate of change of acceleration, the comfort cost of the locus candidate is calculated by, for example, a comfort cost algorithm or a function. Then, the total locus cost can be calculated based on the target cost and the comfort cost, and the target locus can be selected based on the total locus cost.

According to one embodiment, when calculating the comfort cost of the locus candidate, the locus candidate is divided into a plurality of candidate sections. For each of the candidate sections, the section comfort cost of the candidate section is calculated based on the rate of change of the acceleration associated with the candidate section. Then, the comfort cost is calculated by a predetermined algorithm based on the section comfort cost of all the candidate sections. In a specific embodiment, the comfort cost is calculated based on the total section comfort cost of the candidate section (for example, the sum of squares).

FIG. 1 is a block diagram showing a network structure of an autonomous vehicle according to an embodiment of the present application. With reference to FIG. 1, the network structure 100 includes an autonomous vehicle 101 that is communicated and connected to one or more servers 103 to 104 via the network 102. Although one self-driving car is shown, a plurality of self-driving cars can be connected to each other via the network 102 and / or to servers 103-104. The network 102 may be any type of network, such as a wired or wireless local area network (LAN), a wide area network (WAN) such as the Internet, a cellular network, a satellite network or a combination thereof. .. Servers 103-104 may be any type of server or server group, such as, for example, a network or cloud server, an app server, a backend server, or a combination thereof. The servers 103 to 104 may be a data analysis server, a content server, a traffic information server, a map / interest point (MPOI) server, a location server, or the like.

The autonomous driving vehicle is a vehicle that can be arranged as an automatic driving mode, and the vehicle navigates and passes through the environment when there is very little or no input from the driver in the automatic driving mode. Such self-driving cars may be equipped with a sensor system. The sensor system comprises one or more sensors arranged to detect information about the driving environment of the vehicle. The vehicle-related controller uses the detected information to navigate and pass through the environment. The autonomous driving vehicle 101 can travel in a manual mode, an automatic driving mode, or a part of the automatic driving modes.

In one embodiment, the autonomous vehicle 101 includes, but is not limited to, a sensing / planning system 110, a vehicle control system 111, a wireless communication system 112, a user interface system 113, and a sensor system 115. .. The self-driving car 101 may further include some common components found in a normal vehicle, such as an engine, wheels, steering wheel, transmission, and the like. The components can be controlled by the vehicle control system 111 and / or the sensing / planning system 110 using a plurality of communication signals and / or commands. The plurality of types of communication signals and / or commands are, for example, acceleration signals or commands, deceleration signals or commands, steering signals or commands, brake signals or commands, and the like.

The components 110 to 115 can be communicated and connected to each other by an interconnector, a bus, a network, or a combination thereof. For example, components 110-115 can be communicatively connected to each other via a controller local area network (CAN) bus. CAN Bus is a vehicle bus standard designed to allow microcontrollers and devices to communicate with each other in applications without a host computer. This is a message-based protocol originally designed for multiple electrical wiring in automobiles, but it is also used in many other environments.

Here, with reference to FIG. 2, in one embodiment, the sensor system 115 includes one or more cameras 211, a Global Positioning System (GPS) unit 212, an inertial measurement unit (IMU) 213, a radar unit 214, and an optical. Includes, but is not limited to, detection and ranging (LIDAR) units 215. The GPS unit 212 may include a transmitter / receiver. The transmitter / receiver can provide information regarding the position of the autonomous driving vehicle by operation. The IMU unit 213 can detect a change in the position and direction of the autonomous vehicle based on the inertial acceleration. The radar unit 214 can be represented as a system that senses an object in the local environment of an autonomous vehicle using wireless electrical signals. In some embodiments, besides sensing the object, the radar unit 214 can additionally sense the speed and / or direction of travel of the object. The lidar unit 215 can use a laser to sense objects in the environment in which the autonomous vehicle is located. In addition to other system components, the lidar unit 215 may further include one or more laser light sources, a laser scanner and one or more detectors. The camera 211 may include one or more devices that collect images of the surrounding environment of the autonomous vehicle. The camera 211 may be a stationary camera and / or a video camera. The camera may be moved mechanically, for example by mounting the camera on a rotating and / or tilting platform.

The sensor system 115 may include other sensors such as, for example, a sonar sensor, an infrared sensor, a steering sensor, an accelerator sensor, a brake sensor and a radio sensor (eg, a microphone). The radio sensor may be arranged to acquire sound from the environment around the autonomous vehicle. The steering sensor may be arranged to detect the steering angle of the steering wheel, the wheels of the vehicle or a combination thereof. The accelerator sensor and the brake sensor detect the accelerator position and the brake position of the vehicle, respectively. In some cases, the accelerator sensor and the brake sensor may be integrated as an integrated accelerator / brake sensor.

In one embodiment, the vehicle control system 111 includes, but is not limited to, a steering unit 201, an accelerator unit 202 (also referred to as an acceleration unit) and a brake unit 203. The steering unit 201 is used to adjust the direction or direction of travel of the vehicle. The accelerator unit 202 is used to control the speed of the motor or engine, and the speed of the motor or engine is further used to control the speed and acceleration of the vehicle. The brake unit 203 decelerates the vehicle by providing friction and decelerating the wheels or tires of the vehicle. The parts shown in FIG. 2 can be implemented by hardware, software, or a combination thereof.

Returning to FIG. 1, the wireless communication system 112 allows communication between the autonomous vehicle 101 and an external system such as a device, a sensor, or another vehicle. For example, the wireless communication system 112 may directly communicate wirelessly with one or more devices, or wirelessly communicates via a communication network, for example, communicates with servers 103 to 104 via a network 102. You may. The wireless communication system 112 can communicate with other components or systems using any cellular communication network or wireless local area network (WLAN), such as WiFi®. The wireless communication system 112 communicates directly with a device (eg, a passenger's mobile device, display device, speaker in vehicle 101) using, for example, an infrared link, Bluetooth®, or the like. The user interface system 113 may be a part of a peripheral device implemented in the vehicle 101, and includes, for example, a keyboard, a touch panel display device, a microphone, a speaker, and the like.

Some or all of the functions of the autonomous vehicle 101 can be controlled or managed by the sensing / planning system 110, especially when operated in the autonomous driving mode. The sensing / planning system 110 receives information from the sensor system 115, the control system 111, the wireless communication system 112 and / or the user interface system 113, processes the received information, and routes or routes from the start point to the target point. Includes necessary hardware (eg, processor, memory, storage device) and software (eg, operating system, planning and route setting program) to drive the vehicle 101 based on planning and control information after planning. .. Instead, the sensing / planning system 110 may be integrated with the vehicle control system 111.

For example, a user who is a passenger can specify a trip start position and destination, for example, via a user interface. The sensing / planning system 110 acquires trip-related data. For example, the sensing / planning system 110 can acquire location and route information from the MPOI server. The MPOI server may be a part of servers 103 to 104. The location server provides a location service, and the MPOI server provides a map service and a POI for a location. Instead, such location and MPOI information may be locally cached in the non-volatile storage of the sensing / planning system 110.

When the autonomous vehicle 101 moves along the route, the sensing / planning system 110 can also acquire real-time traffic information from the traffic information system or server (TIS). The servers 103 to 104 may be operated by a third party entity. Instead, the functions of the servers 103-104 may be integrated with the sensing / planning system 110. Based on real-time traffic information, MPOI information and location information, and real-time local environmental data detected or sensed by the sensor system 115 (eg, obstacles, objects, nearby vehicles), the sensing / planning system 110 is optimal. By planning the route and driving the vehicle 101 according to the planned route, for example, via the control system 111, it is possible to safely and efficiently reach a predetermined destination.

The server 103 may be a data analysis system that executes a data analysis service for various clients. In one embodiment, the data analysis system 103 includes a data collector 121 and a machine learning engine 122. The data collector 121 collects driving statistical data 123 from a plurality of types of vehicles (autonomous driving vehicles or ordinary vehicles driven by a human driver). The driving statistical data 123 includes information indicating issued driving commands (for example, accelerator command, brake command, steering command) and vehicle responses (for example, speed, acceleration, reduction) collected by vehicle sensors at different timings. Information indicating speed, direction) is included. The driving statistics data 123 contains information describing the driving environment at different timings, such as routes (including departure and destination positions), MPOI, weather conditions and road conditions (eg, slowing on highways, traffic jams, traffic accidents). , Road construction, temporary detours, unknown obstacles, etc.) may be further included.

Based on the driving statistical data 123, the machine learning engine 122 generates or trains a ruleset, an algorithm and / or a model 124 for various purposes. Among them, a training model for determining whether or not the obstacle detected by the ADV sensor is a stop sign or a traffic signal is provided. Rule / algorithm 124 may include algorithms for quaternary and / or quintic polynomials that calculate the traffic rules and trajectories that ADV follows. The algorithm 124 may further include an algorithm that generates a locus candidate as the final locus. The algorithm 124 further includes an algorithm for calculating the target cost, the safety cost, and the comfort cost, and may include an algorithm for calculating the locus cost for selecting the target locus from the locus candidates. Then, the algorithm 124 may be uploaded to the ADV and used to generate a real-time locus of automatic driving.

3A and 3B are block diagrams illustrating an example of a sensing / planning system used with an autonomous vehicle according to one embodiment. The system 300 may be implemented as part of the autonomous vehicle 101 of FIG. 1, including, but not limited to, a sensing / planning system 110, a control system 111 and a sensor system 115. With reference to FIGS. 3A to 3B, the sensing / planning system 110 includes a positioning module 301, a sensing module 302, a prediction module 303, a determination module 304, a planning module 305, a control module 306, a route setting module 307, and a trajectory generation module 308. And locus evaluation module 309, but not limited to them.

Some or all of the modules 301-309 may be implemented in software, hardware or a combination thereof. For example, these modules may be installed in the non-volatile storage device 352, loaded into memory 351 and executed by one or more processors (not shown). Note that some or all of these modules may be communicably connected to some or all of the modules of the vehicle control system 111 of FIG. 2, or may be integrated integrally with them. Some of the modules 301 to 309 can be integrated as an integrated module. For example, the locus generation module 308 and the locus evaluation module 309 may be part of the determination module 304 and / or the planning module 305. The decision module 304 and the planning module 305 may be integrated modules.

The positioning module 301 determines the current position of the autonomous vehicle 300 (eg, using the GPS unit 212) and manages any data regarding the user's trips or routes. The positioning module 301 (called the map / route module) manages any data related to the user's trip or route. The user can specify the start position and destination of the trip by registering via, for example, the user interface. The positioning module 301 communicates with other elements such as the map and route information 311 in the autonomous vehicle 300 to acquire data on trips. For example, the positioning module 301 can acquire position and route information from a position server and a map / POI (MPOI) server. The location server provides the location service, and the MPOI server can be cached as part of the map and route information 311 by providing the map service and the POI of a location. When the self-driving car 300 moves along the route, the positioning module 301 can also acquire real-time traffic information from the traffic information system or the server.

The sensing module 302 determines the sensing to the surrounding environment based on the sensor data provided by the sensor system 115 and the positioning information acquired by the positioning module 301. The sensing information can indicate what a normal driver should sense around the vehicle being driven by the driver. Sensing can be, for example, a lane configuration that employs the form of an object (eg, a straight lane or a curved lane), a traffic light signal, the relative position of another vehicle, a pedestrian, a building, a pedestrian crossing, or another traffic-related sign (eg, a straight lane or a curved lane). It can include a stop sign, a sway sign) and the like.

The sensing module 302 may include the function of a computer vision system or computer vision system that processes and analyzes images acquired by one or more cameras to recognize objects and / or features in the environment of an autonomous vehicle. it can. The objects can include traffic signals, road boundaries, other vehicles, pedestrians and / or obstacles and the like. Computer vision systems can use object recognition algorithms, video tracking, and other computer vision technologies. In some embodiments, the computer vision system can draw an environmental map, track the object, estimate the speed of the object, and so on. The sensing module 302 can also detect objects based on other sensor data provided by other sensors such as radar and / or lidar.

For each object, the prediction module 303 predicts how the object will move in this case. This prediction is executed based on the map and the set of the route information 311 and the traffic rule 312 according to the detection data in which the driving environment is detected at the relevant timing. For example, if the object is a vehicle in the opposite direction and the current driving environment includes an intersection, the prediction module 303 predicts the possibility of the vehicle going straight or curving. If the sensing data indicates that there is no traffic light at the intersection, the prediction module 303 can predict that the vehicle needs to be fully parked before entering the intersection. If the sensing data indicates that the vehicle is currently in the left or right turn lane, the prediction module 303 can predict that the vehicle is likely to turn left or right.

For each object, the determination module 304 determines how the object corresponds. For example, for a particular object (eg, another vehicle on an intersecting route) and metadata describing the object (eg, speed, direction, steering angle), how the determination module 304 responds to the object. Determine (for example, overtaking, giving way, stopping, overtaking). The decision module 304 can make such a decision based on a rule set such as traffic rule or driving rule 312, and the rule set can be stored in the non-volatile storage device 352.

The route setting module 307 is arranged to provide one or more routes or routes from the start point to the end point. For a predetermined trip from the start position (received from the user, for example) to the target position, the route setting module 307 acquires the route and map information 311 and all the routes or routes that can be traveled from the start position to the target position. To confirm. The route setting module 307 can generate a reference line based on a topographic map in which each route from the start position to the target position is determined. The reference line is an ideal route or route that is not interfered with, for example, by other vehicles, obstacles or traffic conditions. That is, in the absence of other vehicles, pedestrians or obstacles on the road, the ADV should follow the reference line precisely or closely. Then, the topographic map is provided to the determination module 304 and / or the planning module 305. The decision module 304 and / or the planning module 305 may include other data provided by other modules (eg, traffic conditions from positioning module 301, driving environment sensed by sensing module 302 and traffic predicted by predicting module 303). Based on the situation), inspect all the routes that can be traveled, select one of the optimum routes, and update it. Depending on the particular driving environment at the timing, the actual route or route for controlling the ADV may be different or close to the reference line provided by the route setting module 307.

The planning module 305 is based on the decisions made for each of the sensed objects, based on the reference lines provided by the route setting module 307, for the autonomous vehicle the route or route and driving parameters (eg, distance, speed and). / Or steering angle) is planned. In other words, for a given object, the decision module 304 decides what to do with the object and the planning module 305 decides what to do. For example, for a given object, the decision module 304 can determine to overtake the object, and the planning module 305 can determine to overtake the object to the left or right. The planning and control data is generated by the planning module 305 and includes information that depicts how the vehicle 300 will move in the next travel cycle (eg, next route / route section). For example, planning and control data can indicate that the vehicle 300 travels 10 meters at a speed of 30 miles per hour (mph) and then changes to the right lane at a speed of 25 mph.

Based on the planning and control data, the control module 306 controls and drives the autonomous vehicle by transmitting an appropriate command or signal to the vehicle control system 111 according to the route or route limited by the planning and control data. To do. The planning and control data can be used to route a vehicle from a first point on a route or route by using appropriate vehicle installation or driving parameters (eg, accelerator, brake and steering commands) at different times along the route or route. Have enough information to drive to two points.

The planning step is performed at multiple planning cycles (also called operating cycles) (eg, every 100 milliseconds (ms) time interval). For each planning cycle or operation cycle, one or more control commands are issued based on the planning data and the control data. That is, every 100 ms, the planning module 305 plans the next route section or route section, and includes, for example, the target position and the time required for the ADV to reach the target position. Alternatively, the planning module 305 can further specify specific speeds, directions and / or steering angles and the like. In one embodiment, the planning module 305 plans a route section or route section for the next scheduled time zone (eg, 5 seconds). For each planning cycle, the planning module 305 plans the target position to be used in the current cycle (eg, the next 5 seconds) based on the target position planned in the previous cycle. Then, the control module 306 generates one or a plurality of control commands (for example, accelerator, brake, steering control commands) based on the planning data and the control data in the current cycle.

The determination module 304 and the planning module 305 can be integrated as an integrated module. The determination module 304 / planning module 305 can be provided with the function of a navigation system or a navigation system so as to determine the driving route of the autonomous driving vehicle. For example, a navigation system can determine a series of speeds and directions of travel that enable an autonomous vehicle to travel along the following routes: In the above-mentioned route, the self-driving car is advanced along the route by the lane for traveling to the final destination, and the detected obstacle is substantially avoided. The destination may be set based on user input made via the user interface system 113. The navigation system can dynamically update the driving route at the same time as the self-driving car is driving. The navigation system can merge data from one or more maps with the GPS system to determine the driving route for the self-driving car.

The decision module 304 / planning module 305 may further include the function of an anti-collision system or anti-collision system for recognizing, evaluating, avoiding or otherwise passing through potential obstacles in the environment of the autonomous vehicle. it can. For example, the collision prevention system operates one or more subsystems of the control system 111 to perform a direction change operation, a curve running operation, a braking operation, and the like to realize a change during navigation of an autonomous vehicle. Can be done. The collision prevention system can automatically determine possible obstacle avoidance actions based on surrounding traffic patterns, road conditions, and the like. The collision prevention system is configured so that when an autonomous vehicle detects a vehicle, building obstacle, etc. in an adjacent area that is about to change direction and enter by another sensor system, it does not change direction. Can be done. The anti-collision system maximizes the safety of the occupants of the self-driving car and can automatically select the available actions. The anti-collision system can select the avoidance action that is expected to generate the least acceleration in the cabin of the autonomous vehicle.

According to one embodiment, the locus generation module 308 can generate longitudinal locus candidates used in the ADV 101. For example, trajectory candidates can be generated in consideration of factors such as safety, comfort and traffic rules. Then, the ADV can be controlled by selecting any one of the locus candidates generated by the cost function 316 as the locus based on the locus evaluation module 309 (also called the locus selection module) evaluation standard 315. The process of trajectory selection will be described in detail below. The locus generation module 308 can be implemented as part of the decision module 304 and / or the planning module 305. In one embodiment, the locus generation module 308 can generate station-time graphs in real time depending on the sensing environment. Then, the locus generation module 308 can project any obstacle detected in the example onto the station-time graph. The end status can be listed based on the boundaries of the projected obstacles. Then, a trajectory candidate can be generated based on these end situations. ADV can be controlled by selecting a trajectory candidate as the final trajectory.

FIG. 4 is a block diagram showing an example of a locus generation module according to an embodiment. With reference to FIG. 4, the locus generation module 308 includes an initial status determination device 401, a scene determination device 403, a figure generation device 405, an end status determination device 407, a trajectory candidate generation device 409, and a trajectory candidate selection device. 411 and may be provided. The initial status determination device 401 can determine the initial status of the ADV. The scene determination device 403 can determine the driving scene of the ADV based on the obstacle detected by the ADV. The graphic generator 405 can generate a station-time graph and / or a speed-time graph of ADV. The termination status determination device 407 can determine all possible termination statuses of the ADV based on the driving scene and the initial status of the ADV, and other factors such as safety, comfort and traffic rules. The locus candidate generation device 409 can generate locus candidates according to possible end situations. The locus candidate selection device 411 can select locus candidates from the locus candidate pool.

FIG. 5 is a block diagram illustrating an example of an ADV scene according to one embodiment. With reference to FIG. 5, scene 500 includes ADV 101 along path 505. The route 505 may be a route (or a section of the route) predetermined by the determination module 304 and / or the planning module 305, and the ADV 101 can be traveled from the start point to the final destination. The ADV101 can sense the surrounding environment including the preceding vehicle 504, the red traffic signal 506 ahead and the vehicle 502 (predictable by the ADV101 prediction module 303 so that there is a chance to steer in front of the ADV101). Is).

With reference to FIGS. 4 and 5, in one embodiment, the initial status determination device (eg, the initial status determination device 401 of the ADV 101) can determine the initial status of the ADV 101 based on the current vehicle state. The vehicle state may include the position, speed, acceleration, curvature (eg, direction of travel) and rate of change of the curvature of the ADV. Here, the initial status determination device 401 can determine the initial (for example, current) status of the vertical position (s_init), vertical velocity (s_dot_init), and vertical acceleration (s_ddot_init) of the ADV 101.

In one embodiment, the figure generator 405 is a vertical speed-time (or speed-time or VT) having a predetermined time zone (ie, the maximum planned time interval of the current planned cycle) based on the initial situation. Generate a graph. FIG. 6A is an example of a speed-time graph according to an embodiment. FIG. 6A may be a speed-time graph generated by the graphic generator 405 and corresponding to the scene 500 of FIG. With reference to FIG. 6A, the VT graph 600 includes a start point 601 and a plurality of endpoints (eg, points 603-605) corresponding to a plurality of possible end situations. In one embodiment, the number of termination situations can be enumerated based on constant time intervals and / or speed increases. In other embodiments, the number of termination situations can be enumerated based on a user-defined time interval. In other embodiments, the number of termination situations can be limited by map information and / or traffic rules (eg, maximum and / or minimum speed limits in road sections).

With reference to FIG. 6A, point 601 can correspond to the initial longitudinal velocity (eg, s_dot_init) at timing t = 0 of ADV101. Point 603 and point 605 can correspond to the two enumerated end situations and can indicate the endpoints of the two possible locus candidates. The point 603 corresponds to the end situation in which the speed becomes zero after a while, and the point 605 can correspond to the end situation in which the speed is higher than the initial vertical speed after a while. The end status generated by the speed-time graph can indicate the first set (or second set) of the end status. Each of the end situations in the first set (or the second set) can correspond to the end longitudinal velocity (s_dot_end) and the end longitudinal acceleration (s_ddot_end). Note that in some embodiments, the end longitudinal acceleration is always equal to zero.

Next, the scene determination device 403 determines the current driving scene of the ADV 101. In one embodiment, possible driving situations include, but are not limited to, cruising, following a vehicle in front, and stopping. The current driving scene can be determined based on the current traffic conditions along Route 505 (eg, by the sensing module 302 and / or the prediction module 303), which is based on the obstacles recognized along Route 505. It can be determined. For example, when there are no obstacles in the planned route of ADV101, the driving scene may be a cruising scene. When the planned route of ADV101 includes static (eg, non-moving) and dynamic (eg, moving) obstacles such as a stop sign and a moving vehicle, the driving scene is to the vehicle in front. It may be a follow-up scene. The driving scene may be a stop scene when the planned route contains only static obstacles such as a stop sign and / or a traffic signal. Here, the scene 500 may include static and dynamic obstacles, and the current driving scene may be determined as a follow-up scene.

In one embodiment, the figure generator 405 generates a station-time graph that is used for follow or stop scenes (but not for cruising scenes). The cruising scene may not be related to the station-time graph (route-time graph or route-time section) because there are no detected obstacles. In one embodiment, the graphic generator 405 generates a route-time (station-time) graph based on the initial situation for the follow-up scene or the stop scene. A station-time graph is generated within a predetermined time zone to take into account possible longitudinal trajectories for ADV with sufficient time until the next planned cycle. In one embodiment, the figure generator 405 generates a geometric region of the obstacles detected by the ADV 101 by projecting these obstacles onto the generated station-time graph. These areas in the route-time interval include projections used for static and dynamic obstacles, such as predicted trajectories of other vehicles (eg, vehicle 502) turned to route 505. Is also good.

FIG. 6B shows an example of a station-time graph according to an embodiment. FIG. 6B may be a station-time graph corresponding to the scene 500 of FIG. 5 generated by the graphic generator 405. With reference to FIG. 6B, for example, point 611 can correspond to the vertical position s_init of ADV101. The figure generator 405 is capable of generating station-time obstacles (eg, regions) 612, 614 and 616. Region 612 can accommodate obstacles (or vehicles) 502 that can be predicted to turn in front of ADV101. The area 614 corresponds to the vehicle 504 cruising in front of the ADV 101, and the area 616 corresponds to the red traffic signal 506.

Next, the end status determination device 407 can enumerate the end status based on the generated station-time graph. In one embodiment, the end status can be enumerated based on each boundary of the region in the station-time graph. In other embodiments, the end status can be enumerated based solely on the upper and / or lower boundaries of each region in the station-time graph. For example, ST points 621-622 along the upper and / or lower boundaries of region 612 can be listed as the end status of region 612. ST points 624 along the lower boundary of region 614 can be listed as the end status of region 614. ST points 626 to 627 along the lower boundary of region 616 can be listed as the end status of region 616. Here, when the ADV101 touches the areas 612, 614, 616 or intersects the areas 612, 614, 616, the ADV101 violates the traffic rules and / or collides with an obstacle, so that the areas 612, 614, 616 There is no need to list the end status. In one embodiment, the number of termination situations can be listed based on a fixed time interval. In other embodiments, the number of termination situations can be enumerated based on a time interval predetermined by the user. The end status generated by the station-time graph can indicate a second set (or first set) of end status. Each of the end situations in the second set (or first set) corresponds to the final station position (s_end), longitudinal velocity (s_dot_end) and longitudinal acceleration (s_ddot_end). Note that in some embodiments, the final longitudinal acceleration is always equal to zero.

In one embodiment, when the second set (or first set) of the end status is determined, the locus candidate generator 409 is from the first set (or second set) and the second set (first set). Fitting these end situations with the corresponding initial situation curves into quaternary (fourth order) and / or fifth order (fifth order) polynomials by using a curve fitting algorithm that achieves smoothness based on the end situation. To generate multiple trajectory candidates. With reference to FIG. 6A (eg, VT graph), for example, the locus candidate generator 409 fits a quartic polynomial to the start point 601 and the endpoint 605 to generate the locus candidate 609 by applying a curve fitting algorithm. can do. In another example, with reference to FIG. 6B (eg, ST graph), the locus candidate generator 409 fits the quintic polynomial to the start point 611 and the endpoint 621 by applying a curve fitting algorithm. Generate 615. In this example, the locus candidate 615 corresponds to the ADV 101 that exceeds the predicted locus of the vehicle 502.

In one embodiment, when locus candidates are generated from all possible end situations, the evaluation module (eg, evaluation module 309 in FIG. 3A) evaluates each of the locus candidates to obtain the optimal locus candidate. Then, the locus candidate selection device 411 can select the optimum locus candidate and control the ADV.

Returning to FIG. 3A, the trajectory evaluation module 309 can be implemented by software, hardware or a combination thereof. For example, the trajectory evaluation module 309 is installed in the non-volatile storage device 352, loaded into memory 351 and can be executed by one or more processors (not shown). The trajectory evaluation module 309 may be communicatively connected to some or all of the modules of the vehicle control system 111 of FIG. 2, or may be integrated with some or all of the modules of the vehicle control system 111 of FIG. .. In one embodiment, the locus evaluation module 309 and the locus generation module 308 may be integrated modules.

FIG. 7 is a block diagram showing an example of a trajectory evaluation module according to an embodiment. With reference to FIG. 7, the locus evaluation module 309 can evaluate the locus candidates based on the cost function so as to select the optimum locus and evaluate the locus candidates for controlling ADV. The locus evaluation module 309 may include a locus partition device 701, a comfort cost determination device 703, a safety cost determination device 705, a reference speed curve generation device 707, and a target / rule compliance cost determination device 709. .. The locus sectioning device 701 can partition the candidate locus into a plurality of sections. The comfort cost determination device 705 can determine the comfort cost of the candidate trajectory. The safety cost determination device 703 can determine the safety cost of the candidate trajectory. The reference speed curve generator 707 can generate a reference speed-time curve (which follows traffic rules) to evaluate the target compliance cost of the candidate trajectory. The target / rule compliance cost determination device 709 can determine the target cost of the candidate trajectory.

Trajectory candidates can be evaluated considering a number of criteria (or costs), such as safety, comfort, rule compliance and / or goal compliance. These criteria may be defined as part of evaluation criteria 315. The safety cost can be calculated by considering whether or not the trajectory candidate guides the automatic vehicle to collide with other vehicles / obstacles on the road. The comfort cost can be calculated in consideration of whether or not the passenger in the ADV has a comfortable riding experience when the locus candidate is used for the ADV. The rule compliance cost can be calculated by considering whether or not the automatic vehicle complies with the traffic rules. It needs to stop, for example, in front of where the vehicle is, such as a stop sign and / or a red traffic light. It is possible to calculate the target compliance cost by considering whether the transportation mission is completed immediately by ADV (that is, whether the trajectory candidate can transport the passenger to the destination without wasting time). it can. In one embodiment, different evaluation criteria can be modeled as cost function 316. The exemplary cost function (or total cost function) may be cost_overall = sum (cost * weight). The cost may be the cost used for comfort, safety, rule compliance and / or goal compliance, and the weight is a weighting coefficient for each of the applicable costs.

With reference to FIGS. 7 and 8, in one embodiment, the safety cost can be evaluated (by the safety cost determination device 703) by the station-time graph (eg, the station-time graph 800 of FIG. 8). The station-time graph 800 may be the station-time graph 610 of FIG. 6B and includes obstacles 612, 614, 616 and trajectory candidates 615 for which safety should be evaluated. In one embodiment, the locus partitioning device 701 partitions locus candidates 615 into a plurality of sections 801 according to a predetermined time resolution. Each interval 801 may be provided with an associated endpoint 803 (eg, a time stamp). Then, the locus partition device 701 defines the endpoint 803 (or time stamp) by the frame 805. The frame 805 may have a predetermined width capable of expressing the width of ADV. In one embodiment, the safety cost determination device 703 has an arbitrary static / dynamic obstacle (eg, obstacles 612, 614, 616) in the station-time graph 800 where any edge of the frame 805 spans each frame 805. Determine if it intersects the boundary of. If there is an intersection (eg, possible collision), the trajectory candidate can be removed from the selection. In another embodiment, the minimum distance from the locus candidate 615 to any boundary of obstacles 612, 614, 616 can be calculated along all the time stamps of the locus candidate 615. The safety cost may be a derivative of the calculated distance, for example, the smaller the distance, the higher the cost. In this case, trajectory candidates away from surrounding obstacles (eg, low cost) may be considered safer than trajectory candidates that may be closer to surrounding obstacles.

The comfort cost can be determined (by the comfort cost determination device 705) by calculating the sum of jerks (for example, the sum of squares) in each time zone. Jerk is defined as a change in the acceleration of ADV. Here, the value of jerk can be calculated from the change in acceleration in the time zone of the adjacent time stamp (for example, endpoint 803). Then, the comfort cost is obtained by obtaining and adding the square value to the value of the jerk in each time zone.

In one embodiment, a reference velocity curve must first be generated in order to determine the target / rule compliance cost (also abbreviated as target cost). The reference velocity is the ideal velocity curve for ADV when there are no obstacles. In other words, the reference velocity curve is a velocity curve that ADV should devise and realize. Without the reference speed curve, the ADV101 does not have a reference speed to compare and is still able to meet the safety and comfort criteria. For example, a stationary ADV101 is safe and comfortable. However, when following the reference speed curve, the ADV 101 can be devised so as to reach a predetermined destination in the shortest time and can ensure comfortable driving.

In one embodiment, there are at least two situations to consider in order to generate a reference velocity curve. The first scene is one in which there are no obstacles and the target of the ADV is cruising at a constant speed, for example v_cruise. In this case, the reference curve generator 707 can generate the reference velocity curve as a constant velocity curve such as the curve 901 of FIG. 9A. In one embodiment, a constant speed or cruising speed is determined based on the road speed limit. In another embodiment, a constant speed or cruising speed is determined based on a speed preset by the ADV user. The reference curve generator 707 may determine a comfortable deceleration d_comf when it is necessary to stop at a point where the ADV has a known position, which is the second scene where there is an obstacle corresponding to the stop point. it can. It may be obtained by multiplying the maximum deceleration d_max of the vehicle by a predetermined comfort coefficient f. The reference curve generator 707 can calculate the deceleration locus (for example, 912 in FIG. 9B) based on d_comf. It can slow down the ADV with a comfortable deceleration (eg, d_comf) and stop the ADV at a known stop point.

In one embodiment, the reference curve generator 707 stops from the current position of the ADV if the ADV is unable to decelerate from the cruising speed v_cruise using the d_comf based on the calculated deceleration trajectory (eg, 912). Generate a constant deceleration curve to a point. The reference velocity curve 921 of FIG. 9C is an example of such a reference velocity curve.

In another embodiment, the reference curve generator 707 has a constant velocity in the first portion where the ADV can be decelerated from the cruising speed v_cruise using d_comf based on the calculated deceleration trajectory (eg, 912). After generating a curve having a locus (eg v_cruise), a constant deceleration locus (eg 912) is generated in the second portion. The reference velocity curve 911 of FIG. 9B is an example of such a reference velocity curve. Once the reference velocity curve is generated for the trajectory candidate, the target / rule compliance cost can be determined.

FIG. 10 is an example showing a speed-time graph of target evaluation according to one embodiment. With reference to FIG. 10, the speed-time graph 1000 includes a speed-time locus 1015 and a reference speed curve 911. The locus 1015 shows the locus candidate 615 of FIG. 8 in the velocity-time domain. The reference speed curve 911 may be a reference speed curve generated for the locus candidate 615 according to the stop point (for example, the red traffic signal 506 in FIG. 5).

In one embodiment, the target / rule compliance cost determination device 709 first has a difference curve (eg, a curve) based on the difference (eg, 1005) from the reference velocity curve 911 at each point (or time stamp) of the locus 1015. 1001) is calculated. Then, the speed value of the difference curve 1005 in each time stamp is added (for example, the sum of squares) to obtain the target / rule compliance cost. Here, the target / rule compliance cost indicates the “proximity” between the trajectory candidate and the reference trajectory.

In one embodiment, the total cost or total cost function may be cost_overall = cost_safety * weight_safety + cost_comfort * weight_comfort + cost_obj * weight, taking into account safety costs, comfort costs, target compliance costs and rule compliance costs. cost_safety is a safety cost, and weight_safety is a weighting coefficient of safety cost. cost_comfort is a comfort cost, and weight_comfort is a weighting coefficient of comfort cost. cost_obj is the target cost, and weight_obj is the weighting coefficient of the target cost. Note that the different weighting factors may be numerical and adjust the importance of each cost to other costs. When the total cost is calculated for each of the locus candidates, the optimum locus candidate can be selected based on the lowest total cost among the locus candidates.

FIG. 11 is a flowchart showing a method executed by ADV according to one embodiment. The flow 1100 can be executed by the processing logic. The processing logic may include software, hardware, or a combination thereof. For example, the flow 1100 can be executed by the locus generation module 308 and / or the locus evaluation module 309 of FIG. 3A. With reference to FIG. 11, in the block 1101, a plurality of locus candidates from the start point to the end point of a specific driving scene are generated for the autonomous driving vehicle (ADV). In block 1102, the processing logic generates a reference locus corresponding to the driving scene based on the current state of the ADV related to the start point and the end state of the ADV related to the end point. Here, the reference trajectory is related to the target. In block 1103, for each of the locus candidates, the processing logic compares the locus candidate with the reference locus and generates a target cost indicating the similarity between the locus candidate and the reference locus. In block 1104, the processing logic selects any one of the locus candidates as the target locus for driving the ADV based on the target cost of the locus candidate.

In one embodiment, comparing a locus candidate with a reference locus means that the locus candidate is divided into a plurality of candidate sections, the reference locus is divided into a plurality of reference sections, and for each of the candidate sections, the candidate section is divided. This includes calculating the speed difference between the speed and the speed of the corresponding section reference section and calculating the target cost based on the speed difference between the candidate section and the reference section. In another embodiment, the target cost is calculated by adding the square of the speed difference between the candidate section and the reference section.

In one embodiment, for each of the locus candidates, the processing logic further recognizes the obstacle closest to the locus candidate, measures the distance between the locus candidate and the obstacle, and sets the locus candidate and the obstacle. The safety cost associated with the locus candidate is calculated based on the distance between, and the total locus cost of the locus candidate is calculated based on the target cost and the safety cost. Then, the target locus can be selected from the locus candidates, and the target locus has the minimum total locus cost.

In one embodiment, for each of the locus candidates, the processing logic further determines the rate of change in acceleration along the locus candidate and calculates the comfort cost of the locus candidate based on the rate of change in acceleration along the locus candidate. , Calculate the total trajectory cost of the trajectory candidate based on the target cost and the comfort cost. Then, the target locus can be selected from the locus candidates, and the target locus has the minimum total locus cost. In another embodiment, calculating the comfort cost of a locus candidate divides the locus candidate into a plurality of candidate sections, and for each of the candidate sections, the candidate section is based on the rate of change in acceleration associated with the candidate section. Includes calculating the section comfort cost of and calculating the comfort cost based on the section comfort cost of the candidate section. In another embodiment, the comfort cost is calculated based on the total section comfort cost of the candidate section.

It should be noted that some or all of the components exemplified and described above can be realized by software, hardware, or a combination thereof. For example, such components may be implemented as software installed and stored in persistent storage, such software as a processor (not shown) to achieve the processes or operations described throughout this application. ) May be loaded into memory and executed. Alternatively, such components are programmed or embedded in dedicated hardware such as integrated circuits (eg, application specific integrated circuits or ASICs), digital signal processors (DSPs), or field programmable gate arrays (FPGAs). It may be implemented as executable code, which can be accessed via the corresponding driver and / or operating system from the application. Moreover, such components can be implemented as specific hardware logic in the processor or processor core as part of a command set in which the software components are accessible by one or more specific commands.

FIG. 12 is a block diagram showing an example of a data processing system that can be used with one embodiment of the present application. For example, system 1500 can represent any of the data processing systems that perform any of the processes or methods (eg, either the sensing and planning system 110 of FIG. 1 or servers 103-104). System 1500 can include a plurality of different components. These components can be implemented as integrated circuits (ICs), parts of integrated circuits, discrete electronic devices, or other modules suitable for circuit boards (eg, motherboards or add-in cards for computer systems). Alternatively, it can be realized as a component incorporated in the chassis of a computer system in another form.

It should be noted that system 1500 is intended to provide a high level view of a plurality of components of a computer system. However, it should be understood that additional components may be present in certain embodiments, and that the components shown in other embodiments can be arranged differently. System 1500 includes desktop computers, laptop computers, tablet computers, servers, mobile phones, media players, personal digital assistants (PDAs), smart watches, personal communicators, gaming devices, network routers or hubs, wireless access points (APs) or It can represent a repeater, a set top box, or a combination thereof. Moreover, although only a single device or system has been shown, the term "equipment" or "system" is used alone or jointly to implement any one or more of the methods described herein. It shall be construed to include any combination of devices or systems that execute one (or more) command set.

In one embodiment, the system 1500 includes a processor 1501, a memory 1503, and devices 1505-1508 that are connected via a bus or interconnect 1510. Processor 1501 can represent a single processor or a plurality of processors including a single processor core or a plurality of processor cores. Processor 1501 can represent one or more general purpose processors such as microprocessors, central processing units (CPUs), and the like. More specifically, processor 1501 executes complex command set computing (CISC) microprocessor, reduced command set computing (RISC) microprocessor, ultralong command word (VLIW) microprocessor, or other command set. It may be a processor or a processor that executes a combination of command sets. Processor 1501 further includes, for example, application-specific integrated circuits (ASICs), cellular or baseband processors, field programmable gate arrays (FPGAs), digital signal processors (DSPs), network processors, graphics processors, communication processors, encryption processors. It may be one or more dedicated processors, such as coprocessors, embedded processors, or any other type of logic capable of processing commands.

The processor 1501 may be a low power multi-core processor socket such as an ultra-low voltage processor and can function as a main processing unit and a central hub for communicating with various components of the system. Such a processor can be implemented as a system on chip (SoC). Processor 1501 is configured to execute commands for performing the operations and steps described herein. System 1500 may further include a graphics interface that communicates with the graphics subsystem 1504 if desired, which graphics subsystem 1504 may include a display controller, a graphics processor, and / or a display device.

The processor 1501 can communicate with the memory 1503, which in one embodiment can be implemented by a plurality of memory devices for providing a predetermined amount of system memory. The memory 1503 is one or more volatile storage (or such as random access memory (RAM), dynamic RAM (RAM), synchronous DRAM (SDRAM), static RAM (SRAM), or other type of storage device. Memory) device can be included. Memory 1503 can store information, including command sequences executed by processor 1501 or any other device. For example, executable code and / or data for various operating systems, device drivers, firmware (eg, basic input / output systems or BIOS), and / or applications are loaded into memory 1503 and executed by processor 1501. Can be done. Operating systems include, for example, Robot Operating System (ROS), Windows® Operating System of Microsoft®, MacOS® / iOS® of Apple, and Google®. It may be any type of operating system, such as Android®, LINUX®, UNIX®, or any other real-time or embedded operating system.

The system 1500 can further include I / O devices such as device 1505 to 1508, which includes a network interface device 1505, an input device 1506 if desired, and other I / O devices 1507 if desired. The network interface device 1505 may include a wireless transmitter / receiver and / or a network interface card (NIC). The wireless transmitter / receiver is a WiFi transmitter / receiver, an infrared transmitter / receiver, a Bluetooth transmitter / receiver, a WiMax transmitter / receiver, a wireless mobile phone transmitter / receiver, a satellite transmitter / receiver (for example, a Global Positioning System (GPS) transmitter / receiver), or other radio frequencies. It may be a (RF) transmitter / receiver or a combination thereof. The NIC may be an Ethernet® card (Ethernet® card).

The input device 1506 was displayed as part of a mouse, touch panel, touch screen (which may be integrated with display device 1504), pointer device (eg, stylus), and / or keyboard (eg, physical keyboard or touch screen). Virtual keyboard) can be included. For example, the input device 1506 can include a touch screen controller that connects to the touch screen. Touch screens and touch screen controllers include, for example, any of a variety of touch sensitive technologies, including but not limited to capacitors, resistors, infrared, and surface sonic technologies, as well as other proximity sensor arrays, or touch. Other elements for determining one or more points of contact with the screen can be used to detect their contact and movement or intermittentness.

The I / O device 1507 can include a voice device. The voice device may include speakers and / or microphones to facilitate voice support functions such as voice recognition, voice replication, digital recording, and / or telephone functions. Other I / O devices 1507 also include universal serial bus (USB) ports, parallel ports, serial ports, printers, network interfaces, bus bridges (eg, PCI-PCI bridges), sensors (eg, accelerometers, etc.). Motion sensors, gyroscopes, accelerometers, optical sensors, compasses, proximity sensors, etc.), or combinations thereof. The apparatus 1507 may further include an imaging processing subsystem (eg, a camera), which is a charge-coupled device (eg,) for facilitating camera functions such as recording photographs and video fragments. Optical sensors such as CCD) or complementary metal oxide semiconductor (CMOS) optical sensors can be included. Certain sensors may be connected to the interconnect 1510 via a sensor hub (not shown), and other devices such as keyboards or thermal sensors may be embedded controllers (figure), depending on the specific arrangement or design of the system 1500. Can be controlled by).

A large capacity storage device (not shown) can be connected to the processor 1501 to provide persistent storage of information such as data, applications, one or more operating systems, and the like. In various embodiments, such mass storage devices can be implemented by solid state devices (SSDs) to improve system responsiveness while enabling thinner and lighter system designs. .. However, in other embodiments, the large capacity storage device can be realized primarily by using a hard disk drive (HDD), and the smaller capacity SSD storage device functions as an SSD cache to prevent a power failure event. In the meantime, non-volatile storage of context states and other such information is possible, which allows for faster energization when system operation resumes. The flash device can also be connected to processor 1501 via, for example, the Serial Peripheral Interface (SPI). Such a flash device can function for non-volatile storage of system software, including the BIOS of the system and other firmware.

The storage device 1508 may include a computer-readable medium 1509 (also referred to as a machine-readable storage medium or a computer-readable medium), wherein the computer-readable medium 1509 is one or more of the ones described herein. One or more command sets or software (eg, modules, units, and / or logic 1528) that embody the method or function of. The processing module / unit / logic 1528 can represent any of the above components, such as the locus generation module 308 of FIG. 3A. The processing module / unit / logic 1528 may further be present entirely or at least partially in memory 1503 and / or in processor 1501 during execution by data processing system 1500, memory 1503, and processor 1501. , Data processing system 1500, memory 1503, and processor 1501 also constitute a machine-accessible storage medium. The processing module / unit / logic 1528 may also be transmitted and received by the network via the network interface device 1505.

Computer-readable storage media 1509 can be used to permanently store some of the software features described above. The computer-readable storage medium 1509 is shown as a single medium in an exemplary embodiment, while the term "computer-readable storage medium" refers to a single medium or a plurality of media in which one or more command sets are stored. It shall be construed to include media (eg, centralized or distributed databases, and / or associated caches and servers). Also, the term "computer-readable storage medium" is construed to include any medium capable of storing or encoding a command set, the command set being executed by a machine and any one or more of the methods of the present application. Is for causing the above machine to execute. Therefore, the term "computer-readable storage medium" shall be construed to include, but is not limited to, solid-state memory, optical and magnetic media, or any other non-transitory machine-readable medium. ..

The processing modules / units / logic 1528, components and other features described herein may be implemented as discrete hardware components or hardware components (eg ASICS, FPGAs, DSPs or similar). It may be integrated into the function of the device). Further, the processing module / unit / logic 1528 may be implemented as firmware or a functional circuit in a hardware device. Further, the processing module / unit / logic 1528 may be realized by any combination of hardware devices and software components.

It should be noted that system 1500 is shown to have various components of a data processing system, but is not intended to represent any particular architecture or method of interconnecting the components, such. The details are not closely related to the embodiments of the present application. It should also be understood that network computers, handheld computers, mobile phones, servers, and / or other data processing systems with fewer or more components can also be used with embodiments of the present application.

Some of the above specific description is already shown in algorithms and symbolic representations of operations on data bits in computer memory. Descriptions and representations of these algorithms are the methods used by those skilled in the art of data processing to most effectively convey their work substance to other skilled in the art. As used herein, the algorithm is generally considered to be a self-consistent sequence that leads to the desired result. These actions require physical treatment of physical quantities.

However, it should be kept in mind that all of these and similar terms are associated with appropriate physical quantities and are only to facilitate labeling of these quantities. Unless otherwise explicitly stated in the above description, what should be understood throughout this specification is that the description in terms (eg, as described in the accompanying patent claims) is a computer system. , Or a similar operation or process of an electronic computing device, the computer system or electronic computing device controls data shown as physical (electronic) quantities in the registers and memory of the computer system, and the data is used in the computer system. Convert to other data also indicated as a physical quantity in a memory or register or such other information storage device, transmission or display device.

Embodiments of the present application also relate to devices for performing the operations herein. Such computer programs are stored on non-temporary computer-readable media. Machine-readable media include any mechanism for storing information in a form readable by a machine (eg, a computer). For example, a machine-readable (eg, computer-readable) medium is a machine (eg, computer) readable storage medium (eg, read-only memory (“ROM”), random access memory (“RAM”), magnetic disk storage medium, optical storage. Medium, flash memory device) included.

The processes or methods described in the drawings above include processing involving hardware (eg, circuits, dedicated logic, etc.), software (eg, those embodied on a non-transitory computer-readable medium), or a combination of both. Can be executed by logic. Although the process or method has been described above in a particular order, it should be understood that some of the operations may be performed in a different order. Also, some operations may be performed in parallel rather than in sequence.

The embodiments of the present application are described without reference to any particular programming language. It should be understood that various programming languages can be used to implement the teachings of the embodiments of the present application described herein.

In the above specification, the embodiments of the present application have already been described with reference to specific exemplary embodiments thereof. As will be apparent, various modifications can be made to the present invention without departing from the broader intent and scope of the present application set forth in the appended claims. Therefore, the specification and drawings should be understood in an exemplary sense, not in a limited sense.

Claims (19)

A computer implementation method that determines the trajectory for operating an autonomous vehicle (ADV).
Steps to generate multiple trajectory candidates from the start to the end of a particular driving scene for autonomous vehicles (ADVs) based on one or more factors of safety, comfort and traffic rules .
A step of based on the end state of the automatic operation vehicle relating to the current state and the end point of the automatic operation wheel associated with the origin, to generate referential locus that corresponds to the target obstacle in the operation scene,
For each of the locus candidates, a step of comparing the locus candidate and the reference locus to generate a target cost indicating similarity between the locus candidate and the reference locus.
Wherein based on the target cost of the path candidates, saw including the steps of: selecting one of said path candidate as a target path for operating the automatic operation vehicle,
The step of comparing the locus candidate and the reference locus is
A step of dividing the trajectory candidate into a plurality of candidate sections, and
A step of dividing the reference locus into a plurality of reference sections, and
For each of the candidate sections, a step of calculating the speed difference between the speed of the candidate section and the speed of the corresponding reference section, and
A method including a step of calculating the target cost based on a speed difference between the candidate section and the reference section . The method according to claim 1 , wherein the target cost is calculated by adding the square of the speed difference between the candidate section and the reference section. For each of the trajectory candidates
Recognize the obstacle closest to the trajectory candidate and
The distance between the trajectory candidate and the obstacle is measured.
The safety cost associated with the trajectory candidate is calculated based on the distance between the trajectory candidate and the obstacle.
Based on the target cost and the safety cost, the total trajectory cost of the trajectory candidate is calculated.
The method of claim 1, further comprising a step of selecting the target trajectory having the smallest total trajectory cost from the trajectory candidates. For each of the trajectory candidates
Determine the rate of change of acceleration along the trajectory candidate,
Based on the rate of change of acceleration along the trajectory candidate, the comfort cost of the trajectory candidate is calculated.
Based on the target cost and the comfort cost, the total trajectory cost of the trajectory candidate is calculated.
The method of claim 1, further comprising a step of selecting the target trajectory having the smallest total trajectory cost from the trajectory candidates. The step of calculating the comfort cost of the trajectory candidate is
The trajectory candidate is divided into a plurality of candidate sections, and the trajectory candidate is divided into a plurality of candidate sections.
For each of the candidate sections, the section comfort cost of the candidate section is calculated based on the rate of change of the acceleration associated with the candidate section.
The method according to claim 4 , wherein the comfort cost is calculated based on the section comfort cost of the candidate section, which includes a step. The method according to claim 5 , wherein the comfort cost is calculated based on the total of the section comfort costs of the candidate section. A non-temporary machine-readable medium in which instructions are stored
When the command is executed by one or more processors, the one or more processors are made to execute the process, and the process is performed.
Generate multiple trajectory candidates from the start to the end of a particular driving scene for use in autonomous vehicles (ADVs) based on one or more factors of safety, comfort and traffic rules .
Based on the end state of the automatic operation vehicle relating to the current state and the end point of the automatic operation wheel associated with the origin, it generates references trajectory that corresponds to the target obstacle in the operation scene,
For each of the locus candidates, the locus candidate and the reference locus are compared to generate a target cost indicating the similarity between the locus candidate and the reference locus.
Based on the target cost of the locus candidate, any one of the locus candidates is selected as the target locus for driving the autonomous driving vehicle.
Look at including it,
Comparing the locus candidate with the reference locus
Dividing the trajectory candidate into a plurality of candidate sections and
Dividing the reference locus into a plurality of reference sections and
For each of the candidate sections, calculating the speed difference between the speed of the candidate section and the speed of the corresponding reference section.
A non-transitory machine-readable medium comprising calculating the target cost based on a speed difference between the candidate section and the reference section . The non-transitory machine-readable medium according to claim 7 , wherein the target cost is calculated by adding the square of the speed difference between the candidate section and the reference section. For each of the trajectory candidates
Recognize the obstacle closest to the trajectory candidate and
The distance between the trajectory candidate and the obstacle is measured.
The safety cost associated with the trajectory candidate is calculated based on the distance between the trajectory candidate and the obstacle.
Based on the target cost and the safety cost, the total trajectory cost of the trajectory candidate is calculated.
The non-transitory machine-readable medium of claim 7 , further comprising a step of selecting the target locus having the lowest total locus cost from the locus candidates. For each of the trajectory candidates
Determine the rate of change of acceleration along the trajectory candidate,
Based on the rate of change of acceleration along the trajectory candidate, the comfort cost of the trajectory candidate is calculated.
Based on the target cost and the comfort cost, the total trajectory cost of the trajectory candidate is calculated.
The non-transitory machine-readable medium of claim 7 , further comprising a step of selecting the target locus having the lowest total locus cost from the locus candidates. The step of calculating the comfort cost of the trajectory candidate is
The trajectory candidate is divided into a plurality of candidate sections, and the trajectory candidate is divided into a plurality of candidate sections.
For each of the candidate sections, the section comfort cost of the candidate section is calculated based on the rate of change of the acceleration associated with the candidate section.
Wherein based on the interval comfortable cost candidate section to calculate the comfortable cost, non-transitory machine readable medium of claim 1 0, including the step. Non-transitory machine readable medium of claim 1 1 for calculating the comfort costs based on total interval comfort cost of the candidate section. It's a data processing system
With one or more processors
A memory that is connected to the one or more processors and stores commands.
When the command is executed by the one or more processors, the one or more processors are made to execute the process, and the process is performed.
Steps to generate multiple trajectory candidates from the start to the end of a particular driving scene, used in autonomous vehicles (ADVs), based on one or more factors of safety, comfort and traffic rules .
A step of based on the end state of the automatic operation vehicle relating to the current state and the end point of the automatic operation wheel associated with the origin, to generate referential locus that corresponds to the target obstacle in the operation scene,
For each of the locus candidates, a step of comparing the locus candidate and the reference locus to generate a target cost indicating similarity between the locus candidate and the reference locus.
Wherein based on the target cost of the path candidates, saw including the steps of: selecting one of said path candidate as a target path for operating the automatic operation vehicle,
The step of comparing the locus candidate and the reference locus is
A step of dividing the trajectory candidate into a plurality of candidate sections, and
A step of dividing the reference locus into a plurality of reference sections, and
For each of the candidate sections, a step of calculating the speed difference between the speed of the candidate section and the speed of the corresponding reference section, and
A system including a step of calculating the target cost based on a speed difference between the candidate section and the reference section . The system of claim 1 3 for calculating the target cost by adding the square of the speed difference between the candidate section and the reference section. For each of the trajectory candidates
Recognize the obstacle closest to the trajectory candidate and
The distance between the trajectory candidate and the obstacle is measured.
The safety cost associated with the trajectory candidate is calculated based on the distance between the trajectory candidate and the obstacle.
Based on the target cost and the safety cost, the total trajectory cost of the trajectory candidate is calculated.
The system of claim 1 3, wherein selecting the target locus, further comprising the step of having the minimum total path cost from the trajectory candidate. For each of the trajectory candidates
Determine the rate of change of acceleration along the trajectory candidate,
Based on the rate of change of acceleration along the trajectory candidate, the comfort cost of the trajectory candidate is calculated.
Based on the target cost and the comfort cost, the total trajectory cost of the trajectory candidate is calculated.
The system of claim 1 3, wherein selecting the target locus, further comprising the step of having the minimum total path cost from the trajectory candidate. The step of calculating the comfort cost of the trajectory candidate is
The trajectory candidate is divided into a plurality of candidate sections, and the trajectory candidate is divided into a plurality of candidate sections.
For each of the candidate sections, the section comfort cost of the candidate section is calculated based on the rate of change of the acceleration associated with the candidate section.
The system according to claim 16 , which includes a step of calculating the comfort cost based on the section comfort cost of the candidate section. The system according to claim 17 , wherein the comfort cost is calculated based on the total of the section comfort costs of the candidate section. It ’s a computer program
A computer program that realizes the method according to any one of claims 1 to 6 , when the computer program is executed by a processor.